The Yen's Broken Speedometer: A Record Defense Bought the Least Time

On the surface it is the simplest story in global markets. The United States pays high interest rates, Japan pays low ones, so money flows out of the yen and no amount of official yen-buying can stop it. When the yen slid past roughly 162 per dollar this week, near its weakest against the dollar since 1986, that was the explanation on every screen.

The tool markets use to size that 'rate gap' is the difference between US and Japanese government-bond yields, the annual return, in percent, on lending to each government. On the 10-year bond that gap is roughly 1.8 percentage points today (a US Treasury near 4.45 percent against a Japanese Government Bond near 2.7 percent, a yield that has climbed sharply this year as markets bet the Bank of Japan will tighten). In 2022 the same gap was close to 4 points. The long-rate gap that supposedly makes intervention hopeless has more than halved.

And yet Tokyo's April-May 2026 intervention, at ¥11.73 trillion, about $73.6 billion, the largest on record, bought a bounce that faded within a few trading days. In 2022, a smaller spend into a wider gap held for weeks. That inversion is the puzzle.

Japan intervenes rarely and in bursts. In September and October 2022 the Ministry of Finance (MOF), the ministry that funds these operations, spent a combined ¥9.18 trillion, ¥2.83 trillion in September and ¥6.34 trillion in October. In late April and early May 2024 it spent about ¥9.79 trillion (roughly $62 billion). Then, over 28 April to 27 May 2026, it confirmed a record ¥11.73 trillion, about $73.6 billion.

The 2026 defense triggered near 160 yen per dollar, pushed the yen up toward the mid-150s, and then unwound, with the currency printing fresh multi-decade lows within days and touching about 162.6 this week. On the back of that, Reuters reported that Tokyo had shifted to 'ambush' intervention, buying with no advance warning, on the theory that telegraphing its moves let traders betting against the yen close out cheaply.

That policy pivot is why this is worth investigating now. It is a fix for a timing problem. Our question is whether timing is the problem at all.

Interventions almost always jolt the price on the day. The question that matters is whether the jolt survives. So we borrowed a term from physics: half-lifehalf-lifeBorrowed from physics: the number of trading days until the yen has given back half of the bounce an intervention produced. A longer half-life means the defense held., the number of trading days until the yen gives back half of its post-intervention bounce. A long half-life means the market respected the line. A short one means traders bought the cheaper dollar back within hours and kept going.

To keep the four episodes comparable we set 'day zero' at each episode's single largest operation, the biggest one-day yen purchase in that burst, taken from MOF's disclosures, and then tracked the daily close of USD/JPYUSD/JPYThe price of one US dollar in yen; a higher number (for example 162) means a weaker yen. until it had reclaimed half the move from that day's high. Dating the clock to the largest shot, rather than the first, avoids crediting a small opening probe with a bounce that a later, bigger operation actually produced. It is a judgment call, and a different convention would move the day counts by a day or two; it would not flip the ranking.

Line the four episodes up and the simplest read is seductive: spending rose, the long-rate gap fell, and durability collapsed. But the same table that tells that story also refutes the tidy version of it.

Look at 2022. September and October were weeks apart, into a near-identical rate gap of about 4 points, yet the September shot faded in roughly two and a half days while October held for about ten. Same gap, four-fold difference in durability. Whatever separated them, it was not rates. October was more than twice the size and, crucially, was fired after a blow-off topblow-off topA final, steep spike in a price move driven by one-sided crowd positioning, which often leaves the market primed to snap back the other way. near 151.9, into stretched, one-sided positioning that was primed to reverse. Size and timing, not the yield gap, look like the difference.

The carry trade, moreover, is funded at short rates, not the 10-year, so the honest test uses the 2-year gap. That gap tells an even less convenient story for the rate thesis. It narrowed far less than the 10-year (from roughly 4.3 points in 2022 to about 2.9 today), and it was actually at its widest in 2024, at around 4.6 points. Yet 2024's defense held longer than September 2022's did into a narrower front-end gap. Whichever maturity you pick, the rate gap simply does not sort the four defenses from most durable to least. The variable that does the ordering work is not on the yield screen at all.

Two shots weeks apart, into the same rate gap, differed four-fold in how long they held. Whatever set that difference, it was not the carry trade.

Start with the single most telling number, because it rules out the easy escape hatch. If the yen were simply the victim of an unusually strong dollar, it would be weak against the dollar and roughly normal against everyone else. It is not. Adjusted for inflation and measured against a basket of trading partners, what economists call the real effective exchange rate, the yen sits near its weakest in over half a century, the lowest in Bank for International Settlements data that runs back to 1970. The yen is cheap against almost everybody. That is a yen story, not a dollar story.

The candidate explanation is a set of flows that leave Japan every day and do not much care what the exchange rate is, the opposite of a carry tradecarry tradeBorrowing in a low-rate currency like the yen and parking the money in a higher-rate currency like the dollar to pocket the difference; it pushes the borrowed currency down., which is fast, leveraged and reverses the moment the rate gap closes. Three strands stand out. Japan still runs a current-account surpluscurrent-account surplusA measure of whether a country earns more from the rest of the world than it pays out, across trade, investment income and other flows., but a growing share of it is 'primary incomeprimary incomeMoney a country earns on its overseas investments (dividends, interest, factory profits) rather than from selling goods; much of Japan's is reinvested abroad and never converted back into yen.', dividends, interest and factory profits earned abroad, that Japanese companies increasingly leave and reinvest overseas rather than converting back into yen. Those earnings never become yen demand.

The second strand is newer and grows every year: Japan's digital-services deficitdigital-services deficitThe gap between what Japan pays foreign tech firms for cloud, software and ad services and what it earns back, a steady stream of yen sold for dollars., the money it pays foreign tech firms for cloud computing, software and online advertising. That bill hit a record of about ¥6.5 trillion in 2024, roughly 1 percent of the entire economy, and is a stream of yen sold for dollars that keeps flowing whether the yen is at 140 or 162.

The third strand is households themselves. Since Japan supercharged its tax-free NISANISAJapan's tax-free personal investment accounts; expanded in 2024, they have channeled household savings into foreign stocks, which means selling yen to buy them. investment accounts in 2024, ordinary savers have been steadily moving money into foreign, largely US, stocks. Buying a US index fund means selling yen for dollars, month after month, on autopilot. None of these three flows is watching the rate gap. Together they form a persistent, price-insensitive bid for foreign currency that an intervention, however large, meets head-on and then exhausts.

This is what makes Tokyo's new tactic worth scrutinising. An 'ambush', buying with no warning, is designed to punish speculators who had learned to front-run telegraphed interventions and pocket the bounce. Against a purely speculative, leveraged short, that logic is sound: surprise raises the cost of the bet and can force a sharp, painful unwind.

But if a meaningful part of what is pushing the yen down is not a speculative short at all, if it is reinvested corporate profits, a swelling cloud-computing bill and a nation of savers dollar-cost-averaging into foreign funds, then surprise does not touch the driver. A company leaving its Texas profits in dollars is not a trader who can be scared into covering. It has no position to close. The ambush changes who loses money on any given day, from telegraphed speculators to caught-flat-footed ones, without changing the direction of the underlying tide.

You cannot ambush a flow. A saver buying a US index fund every month has no short position to squeeze.

That is the sense in which it may be a fix for the wrong problem. It treats yen weakness as a timing-and-positioning game the authorities can win with better tradecraft, when the evidence from four defenses is that size and the state of the market matter more than the rate gap, and the real effective exchange rate says the pressure is broad and structural. Tactics can make a given intervention land harder. They cannot convert a continuous outflow into a continuous inflow.

None of this proves the carry trade is dead. Four interventions are four data points, not a regression, and rates still matter at the margin, a genuine Bank of Japan tightening cycle that keeps lifting JGB yields would narrow the gap further and could yet do more for the yen than any defense has. The claim here is narrower and, we think, well supported: across these four episodes the rate gap does not order how long the defenses held, and the yen's weakness against a whole basket of partners points to flows the yield screen does not capture.

The stakes are not abstract. A structurally weak yen is, for Japanese households, a standing tax on everything imported, energy, food, the foreign cloud services their employers rent. Each intervention spends real reserves to buy a bounce that fades in days, and the new ambush tactic mainly changes which traders absorb the loss. The bill for the weak yen, meanwhile, keeps arriving at the same address: the Japanese consumer. Until the flows turn, the speedometer will keep reading 'rate gap' while the car is being pushed by something else entirely.